The short answer

Local discovery has changed shape. A customer used to get a map and a list, and you competed for a place on it. Increasingly they get a recommendation, sometimes one, and it is assembled from your Google Business Profile, your reviews, your website and whatever else mentions you. The businesses that get named are the ones whose facts are consistent everywhere, whose services are stated in the words customers use, and whose reviews are recent and specific.

Nearest no longer wins by default. Matching the conditions in the question does.

What the assistant is actually doing

When someone asks for a plumber who can come today, a dentist good with nervous patients, or a design studio that has worked with jewellery brands, the system is not sorting by distance. It is searching, reading and comparing, then synthesising a recommendation, in much the way we described in query fan-out.

That means the question contains conditions, and your job is to be the business that visibly satisfies them. "Open now" is answered by your hours being accurate, including holidays. "Does emergency callouts" is answered by those words existing on your profile and your site. "Good with nervous patients" is answered by your reviews saying so, in customers' own language, recently.

The system reads all of this in prose. It does not reward keyword stuffing, and it cannot infer a service you have never written down.

The four sources that decide the answer

Source What it settles Most common failure
Google Business ProfileExistence, category, hours, location, servicesHours wrong, services list empty, category too generic
Reviews and repliesReputation and what you are actually good atRecent reviews absent, no owner replies, complaints unanswered
Your websiteDetail, pricing, proof, service areasServices described in brand language nobody searches
Third-party mentionsCorroborationOld addresses and dead phone numbers still live on directories

Contradictions are the quiet killer. Two addresses, three phone numbers, one set of hours on the profile and another in your footer: a machine resolving those either picks wrong or picks someone else. Fixing consistency is unglamorous and it is the highest-return hour in local search.

What to fix, in order

Work down this list. Most local businesses find the first three are enough to change what assistants say about them.

  • Make the profile complete and true. Correct category, every service listed explicitly, accurate hours including holiday exceptions, service areas, and current photographs. Empty fields are not neutral; they are missing evidence.
  • Get consistency to one version. Name, address, phone and hours identical across your profile, website, invoices and every directory you can find. Kill listings you no longer control if you can.
  • Restart reviews, and reply. Ask recent customers, specifically, after a job went well. Reply to everything, including the bad ones — the reply is read as evidence of how you handle problems, which is exactly the argument in responding to negative reviews.
  • Give each service its own page. One page per service, in plain language, with what it costs or what it depends on, who it suits, and what happens next. A single services page listing twelve things answers none of the twelve questions.
  • Add the structured data. LocalBusiness details, opening hours, service areas and FAQ markup remove ambiguity for machines. The approach is in structured data for AI search.
  • Publish the proof. Named projects, photographs of real work, recognisable locations. This is what distinguishes you from three competitors with identical claims.

The local pages worth writing

Location pages have a bad reputation because of the version that swaps a city name into an identical template thirty times. That is spam, and it now reads as spam to machines as well as people.

The version that works is genuinely local: the areas you actually serve, the jobs you have done there, travel or callout terms, and anything specific to that place — building types, regulations, parking, seasons. If you cannot write a page about a location that a competitor could not publish unchanged, do not publish it. The same logic applies to industry pages, which is why our own Delhi page talks about how projects run there rather than repeating a service list with a city bolted on.

How to measure something that has no ranking

There is no position number in an assistant's recommendation, so measure the outcome and the mention. Three things worth tracking monthly:

  • Prompt testing. Twenty real customer questions, run across the assistants your customers use, from inside your service area. Record whether you are named, what is said, and whether it is accurate.
  • Profile actions. Calls, direction requests and website clicks from your Business Profile, watched as a trend rather than a monthly number.
  • Source of enquiry. Ask every new enquiry how they found you and keep the answers. It is cruder than analytics and, for assistant-driven discovery, more reliable — for the reasons set out in tracking AI referral traffic.

If the answers about your business are wrong, the fix is almost always a fact you never published, or one you published in three different versions. That is a good problem: it is cheap to fix, and most of your competitors have not noticed theirs. If you want us to run the audit and tell you what the assistants currently say about you, get in touch.